Mantle

Define AI workflows as YAML. Deploy like infrastructure.

Terraform for workflow automation, with native AI. Single binary. Postgres only. Bring your own keys.

name: fetch-and-summarize
description: Fetch data from an API and summarize it with an LLM

inputs:
  url:
    type: string
    description: URL to fetch

steps:
  - name: fetch-data
    action: http/request
    params:
      method: GET
      url: "{{ inputs.url }}"

  - name: summarize
    action: ai/completion
    credential: my-openai-key
    params:
      model: gpt-4o
      prompt: "Summarize: {{ steps['fetch-data'].output.body }}"
      output_schema:
        type: object
        properties:
          summary:
            type: string
          key_points:
            type: array
            items:
              type: string

Why Mantle?

• IaC Lifecycle -- validate / plan / apply / run. Version every workflow definition. Pin executions to immutable versions. Diff before you deploy.
• Single Binary -- One Go binary, one Postgres database. No message queues, no worker fleets, no cluster topology.
• BYOK (Bring Your Own Keys) -- Your API keys live in your database, encrypted with your encryption key. Mantle never proxies through a hosted service. OpenAI, Anthropic, Bedrock, Azure, or self-hosted.
• AI Tool Use -- Multi-turn function calling out of the box. The LLM requests tools, Mantle executes them via connectors, feeds results back. Crash recovery included.
• Checkpoint and Resume -- Steps checkpoint to Postgres after completion. Crash mid-workflow? Restart picks up from the last completed step.
• Secrets as Opaque Handles -- Credentials are AES-256-GCM encrypted at rest and resolved at connector invocation time. Never exposed in expressions, logs, or step outputs.

How Mantle compares to Temporal, n8n, LangChain, Airflow, and others.

Quick Start

# 1. Install
go install github.com/dvflw/mantle/cmd/mantle@latest

# 2. Start Postgres and initialize
docker-compose up -d
mantle init

# 3. Apply your first workflow
mantle apply examples/hello-world.yaml

# 4. Run it
mantle run hello-world

17 example workflows are included in examples/. See the Getting Started guide for a full walkthrough.

Connectors

9 built-in connector actions ship with the binary:

Connector	Actions	Description
HTTP	http/request	REST APIs, webhooks, any HTTP endpoint
AI (OpenAI)	ai/completion	Chat completions, structured output, tool use
AI (Bedrock)	ai/completion	AWS Bedrock models with region routing
Slack	slack/send, slack/history	Send messages, read channel history
Email	email/send	Send via SMTP, plaintext and HTML
Postgres	postgres/query	Parameterized SQL against external databases
S3	s3/put, s3/get, s3/list	Put, get, list objects (S3-compatible)

Need something else? Write a plugin. Any executable that reads JSON from stdin and writes JSON to stdout -- Python, Rust, Node, Bash.

CLI Reference

Workflow Lifecycle

Command	Description
mantle validate <file>	Offline schema validation with line-number errors
mantle plan <file>	Show diff of what will change
mantle apply <file>	Store versioned workflow definition
mantle run <workflow>	Trigger execution (pinned to current version)
mantle cancel <id>	Cancel a running execution
mantle logs <id>	View step-by-step execution history
mantle status <id>	View current execution state

Server and Triggers

Command	Description
mantle serve	Start persistent server with cron and webhook triggers

Authentication and Secrets

Command	Description
mantle secrets create	Create an encrypted credential
mantle secrets list	List stored credentials
mantle secrets delete <name>	Delete a credential
mantle login	Authenticate via API key, auth code PKCE, or device flow
mantle logout	Clear cached credentials

Administration

Command	Description
mantle init	Run database migrations
mantle migrate [status|down]	Migration management (status, rollback)
mantle teams	Manage teams
mantle users	Manage users
mantle roles assign	Assign roles (admin, team_owner, operator)
mantle audit	View audit log
mantle plugins	Manage plugins
mantle library	Shared workflow library
mantle cleanup	Clean up old executions and artifacts
mantle version	Print version info

See the CLI Reference for full usage and flags.

Configuration

Mantle reads configuration from mantle.yaml, environment variables, and CLI flags. A config file is optional -- sensible defaults work with the docker-compose setup out of the box.

database:
  url: postgres://mantle:mantle@localhost:5432/mantle?sslmode=disable

api:
  address: ":8080"

log:
  level: info

Override precedence (highest to lowest): CLI flags > environment variables (MANTLE_ prefix) > config file > defaults.

Architecture

cmd/mantle/          CLI entrypoint (Cobra)
internal/
  cli/               Command definitions
  config/            Config loading, env var overrides
  engine/            Step execution, checkpoint logic, DAG scheduler
  workflow/          YAML parsing, JSON Schema validation, CEL
  connector/         Built-in connectors (HTTP, AI, Slack, Email, Postgres, S3)
  plugin/            JSON stdin/stdout plugin protocol
  secret/            Credential storage, AES-256-GCM encryption, cloud backends
  auth/              API key and OIDC/SSO authentication, RBAC
  api/               REST API server
  server/            Server mode, cron scheduler, webhook ingestion
  db/                Database migrations and queries
  cel/               CEL expression evaluation
  library/           Shared workflow library
  metrics/           Prometheus metrics
  logging/           Structured logging
  audit/             Audit event emission
  version/           Build version info

Key design decisions:

• Postgres only -- No ORM, no premature storage abstractions. Direct SQL queries.
• CEL expressions -- Google's Common Expression Language for data passing between steps.
• JSON plugins -- Custom connectors run as subprocesses communicating via JSON over stdin/stdout.
• Audit from day one -- Every state-changing operation emits an audit event.

Development

git clone https://github.com/dvflw/mantle.git
cd mantle
docker-compose up -d
make build
./mantle version

make build      # Build binary
make test       # Run tests
make lint       # Run golangci-lint
make run        # Run without building (go run)
make dev        # Start docker-compose services
make migrate    # Run database migrations
make clean      # Remove built binary

Contributing

Contributions are welcome. Please open an issue first to discuss what you want to change. See CONTRIBUTING.md for details.

License

Source-available under the Business Source License 1.1. Production use is permitted; commercial resale as a workflow-as-a-service is not. Converts to Apache 2.0 on 2030-03-22.

---

Built with YAML and conviction.